| tags: | |
| - generated_from_trainer | |
| metrics: | |
| - bleu | |
| model-index: | |
| - name: lg-en-v3 | |
| results: [] | |
| <!-- This model card has been generated automatically according to the information the Trainer had access to. You | |
| should probably proofread and complete it, then remove this comment. --> | |
| # lg-en-v3 | |
| This model is a fine-tuned version of [AI-Lab-Makerere/lg_en](https://huggingface.co/AI-Lab-Makerere/lg_en) on an unknown dataset. | |
| It achieves the following results on the evaluation set: | |
| - Loss: 0.9289 | |
| - Bleu: 32.5138 | |
| ## Model description | |
| More information needed | |
| ## Intended uses & limitations | |
| More information needed | |
| ## Training and evaluation data | |
| More information needed | |
| ## Training procedure | |
| ### Training hyperparameters | |
| The following hyperparameters were used during training: | |
| - learning_rate: 4.4271483249908667e-05 | |
| - train_batch_size: 14 | |
| - eval_batch_size: 6 | |
| - seed: 42 | |
| - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 | |
| - lr_scheduler_type: linear | |
| - num_epochs: 2 | |
| - mixed_precision_training: Native AMP | |
| ### Training results | |
| | Training Loss | Epoch | Step | Validation Loss | Bleu | | |
| |:-------------:|:-----:|:----:|:---------------:|:-------:| | |
| | No log | 1.0 | 26 | 1.0323 | 32.6278 | | |
| | No log | 2.0 | 52 | 0.9289 | 32.5138 | | |
| ### Framework versions | |
| - Transformers 4.21.2 | |
| - Pytorch 1.12.1+cu113 | |
| - Datasets 2.4.0 | |
| - Tokenizers 0.12.1 | |